Intelligent Classification System for Visual Inspection of Casting Products Using CNN

This research study presents a thorough method for categorizing visual examinations of casting products with a Convolutional Neural Network (CNN). The methodology entails a comprehensive gathering and preparing a dataset comprising casting product photographs. This is done to ensure consistency and precision in training the model. The study outlines the CNN architecture, specifically built to efficiently catch and acquire complex patterns and characteristics from the images. The model attains an outstanding accuracy rate of 99% by undergoing intensive training, highlighting its strong and dependable capacity to classify casting products by visual inspection. The performance measures, including Precision, Recall, and F1-Score, are computed for the “Ok” and “Defective” categories. The results consistently demonstrate high values, indicating the model's strong capability to accurately differentiate between the two classes. The results highlight the ability of CNNs to automate and improve quality control procedures in production, offering more efficient and dependable inspection solutions for the sector.